The Agent Skills Directory

[PROMPT_INJECTION]: Automated detectors flagged instruction sequences in references/testing.md and agents/test-writer.md as attempts to override system prompts. These are identified as false positives, as the text consists of documentation and test cases specifically designed to teach agents how to identify and resist prompt injection attacks.
[DATA_EXFILTRATION]: A placeholder URL https://auth.your-server.com in agents/auth-upgrader.md was flagged as blacklisted by automated scanners. In the context of this skill, it is used as a generic example value for generating OAuth discovery metadata templates and does not represent an actual exfiltration attempt by the skill.
[PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface because it reads and processes untrusted files from external codebases and can perform sensitive operations via Bash, Write, and Edit tools. This risk is mitigated by a mandatory user confirmation step before the skill dispatches agents to perform automated transformations.
[COMMAND_EXECUTION]: The skill uses the Bash tool to inspect the local environment and execute transformation plans. While these actions are aligned with the skill's purpose of project scaffolding and auditing, the broad scope of the tool represents a security risk if the skill were to process malicious codebase inputs without human oversight.

surface